At manufacturing, construction, transport, and other worksites, alcohol check results taken before a shift or before work begins are sometimes written in a paper log. The log usually records information that identifies the person checked, the date and time of the check, the test result, and the name or mark of the person who filled it in. Digitizing this log does not end with scanning the paper and saving it as a file. The task is to link each check entry on paper accurately to one digital record, and to make sure the original can be found again and reviewed when needed.
Logs kept for a long time tend to include cases that need input rules, such as faded handwriting, inconsistent date formats, blank unit fields, and repeated tests of the same person. That is why matching total entry counts alone does not prove nothing was missed. Fix the scope of the migration first, set aside anything the original alone cannot confirm, and compare counts and fields separately. Only then can you find both missing records and records linked to the wrong entry.
Start by listing what will be migrated
Before entering data, make a list of each log bundle with its period covered, number of volumes, number of pages, and number of check entries on each page. Do not leave a volume off the list even if it mixes several months' forms or has blank pages. The list becomes the reference for deciding later whether a missing page is a real omission or was blank from the start.
Link the location of the original to the digital record as well. For example, you can store the log bundle name and page number with each record, or, if marking the originals is difficult, keep the link information in a separate reconciliation sheet. When a date or result in the digital system looks questionable, you need to know which page of which paper log to check, so you can tell an input error apart from what was originally written in the log.
| Migration stage | What to compare | What to keep |
|---|---|---|
| Before starting | Log bundles, period covered, page counts, entries per page | Migration scope list and location of originals |
| During entry | Person identifier, date and time, result, unit | Entered values plus "illegible" and "pending confirmation" flags |
| After entry | Total count, sample of originals, corrected and unconfirmed items | Reconciliation results and follow-up list |
Reconcile counts and content in different ways
Comparing the number of paper entries on each page with the number of digital entries finds missing lines first. If a page has 20 entries on paper but 19 in the system, a likely omission is clear. But even when counts match, errors can remain, such as copying another person's record or saving a March record under an April date.
So after reconciling counts, compare the person identifier, check date and time, result, and unit against the original. If it is hard to check every record to the same depth, choose a sample that evenly covers log bundles and periods. However, if the sample turns up an error, it is better not to stop at the sample for that range but to review every record entered the same way.
As a hypothetical example, suppose that after migrating six months of logs, you find one entry in a May sample where the month and day of the date were swapped. The observed fact is a single date mismatch, but it needs to be read as a possible sign that the error repeated within the same file bundle or the same input method. A possible action is to recheck all of May, and every range entered the same way, against the originals. You can treat the recheck as finished when the dates, persons, and results in that range match the originals and the corrected entries are kept separate from those not yet confirmed.
Leave unreadable information as a gap
If you enter a plausible guess for a faded digit or a missing unit, the digital system gains information that was never on the paper. If a number cannot be read, mark it "illegible." If the result is visible but the unit is missing, mark it "unit unconfirmed," or use a similar flag that shows the state of the original. This is not a sign of unfinished input. It is a way to keep content that the original alone cannot confirm from mixing with other records.
If you later confirm the content from other materials or the original author's records, record the location of the original log, the material used to confirm it, and the content before and after the correction. If you simply delete the unconfirmed flag, it becomes hard to explain later why the value changed. Also test-enter some records and compare them with the originals to see whether date formats or leading digits of person ID numbers are converted automatically during input, and then set rules accordingly.
Do not automatically remove duplicates when the same person appears several times on the same day. They may be repeat tests, or they may be separate checks. If the time, order, and remarks in the original are not enough to confirm how the entries relate, it is more appropriate to keep each record and note why the relationship could not be confirmed, rather than merging them into one.
Recheck along the conditions that caused the error
Fixing one error does not end the migration problem. Narrow down the cause, such as a bundle scanned out of order, a range where the form changed, or a problem that occurs only with a particular input method, so you can set the scope of the recheck. For example, if values went missing only on a form where the unit field sits in a different place, it is efficient to group the pages that used that form and reconcile them again.
Link each corrected digital record to its original log bundle, page, and reason for correction. This trail is part of the process for fixing digital records, and it also gives a path for review if questions arise later about a result or date. In the migration completion summary, list separately the total number of entries, the range reconciled against the originals, the "illegible" and "unit unconfirmed" items, and the corrected items.
Set access scope for originals and migrated files
Being able to search records in the system does not mean the paper originals can be discarded right away. Check separately whether the originals need to be kept longer, whether the planned reconciliation is complete, and whether the digital records let you find the location of each original. During the period when both originals and digital records exist, decide who may correct which records, to reduce the chance of conflicting content.
For data that can be linked to individuals, such as alcohol check results, the scans and temporary files created during migration also need to be managed. Article 29 of Korea's Personal Information Protection Act requires personal information controllers to take the technical, managerial, and physical measures needed to ensure security, such as establishing an internal management plan and keeping access records. Accordingly, you can limit access to the raw data to those who need it for their work, and confirm whether copies used by external migration workers are returned or deleted after the work ends.
Reports can contain only the aggregates they need, such as monthly check counts or the number of unconfirmed entries, while individual raw data is managed within a separate scope. It is better to set the completion point of the migration not at the moment the files are created, but at the point when the planned reconciliation against the originals is done and you can explain the status of missing, corrected, and unconfirmed records separately.
